Professional Documents
Culture Documents
Editors: P.N. Tandon, R.C. Tripathi et al., pp. © 2011 Nova Science Publishers, Inc.
Chapter 10
ABSTRACT
NeuroEconomics is a rapidly developing multidisciplinary research area that
employs neuroscience techniques to explain economic theories of human behaviour and
makes efforts to bring researchers from the disciplines of psychology, neuroscience, and
economics to a common platform. The foundations of this science lie in the pioneering
experiments of Kahneman and Tversky (1979). The past decade has seen a remarkable
effort by neuroscientists and economists in applying neurophysiological (animal
experiments recoding single cell behaviour) and neuroimaging (such as event related
potentials, functional magnetic resonance imaging) techniques to understand brain
activity while experimental tasks that involve economic decisions are being performed. In
this review we would like to present studies from single cell recording to behavioural
level while presenting various components related to the process of individual decisions.
Along with the future scope of computational modelling, the existing models related to
individual decision making will be presented. Taking further, the upcoming research area
of collective decision making will be reviewed.
making from neuronal level to neuroimaging studies, including some evidence from lesion
studies. In this section, we briefly describe methodological aspects of studying brain function.
The primary information processing cells in the brain are neurons; around 100 billion
neurons can be found in the human brain. The other type of cells, the Glia, outnumber the
neurons by at least 10 times, and are involved in maintaining the structural integrity of brain
tissue and energy metabolism. The dendrites of the neurons gather inputs from other neurons
through an electrochemical junction called the synapse. Information is processed by the
neuron in the form of electrical impulse (action potential) that travels down its tail, the axon,
it is then passed on to the next synapse by releasing a neurotransmitter (Kandel, Schwartz,
Jessell, 2000; Gazzaniga, Ivry, Mangun, 2002). Excitatory transmitter (e.g., glutamate)
increase firing in the post-synaptic neuron, while Inhibitory transmitters (e.g., GABA)
decreases firing in the post-synaptic neuron. Neurophysiological recordings of neural
responses can be done extra-cellularly while awake-animals engage in behavioural tasks. The
invasiveness of neurophysiological recordings (implanting of electrodes) limits its use in
humans. However, it would be possible to record from patients with electrodes implanted for
therapeutic purpose (e.g. epileptic patients).
Brain imaging techniques provide an excellent tool for studying human brain from
network level to systems level. It is possible to non-invasively image brain activity in healthy
volunteers by measuring electrical/magnetic fields created by electrical activity of the brain.
The electrical impulses sent out by the neurons propagate through the brain tissue to the scalp.
Electrodes placed on the scalp can record this current. This is known as Electroencephalo-
gram (EEG). The electrical field also creates a small magnetic field that can be measured
using magneto-encephalogram (MEG). The advantage of EEG and MEG is high temporal
resolution essentially being able to measure signals changing over milliseconds. While
EEG/MEG can be regarded as relatively direct measures of brain activity, the main drawback
of these techniques is the low spatial resolution. It is possible to localize brain activity within
>1 cm. The various currents in the scalp are added up and there is no unique solution to
localization of the signal source inverse problem. MEG is insensitive to signals from radially-
oriented sources and hence does not record from deep brain regions (Volkow, Rosen, Farde,
1997).
The alternative approach to studying function of human brain is to measure blood flow,
which is an indirect measure of neural activity. In a typical positron emission tomography
(PET) study, subject is injected with a radioactive tracer that has a relatively short half-life
(~120 seconds). The subject performs a behavioural task, while the photons emitted from
radioactive decay are counted by the scanner for every location in the brain. Increased blood
flow results in increased delivery of the radioactive tracer, which further results in a larger
signal being emitted from active regions of the brain. PET can be used with radio-labelled
neurotransmitters or receptor agonists, which allows imaging of the level of various
neurotransmitters or receptors in selected regions of the brain. The disadvantages of the PET
technique, apart from the subject being exposed to radioactive substance, are low spatial
(~1cm) and temporal resolution (~2 min).
Functional magnetic resonance imaging (functional MRI or fMRI), has been widely
popular as a brain imaging technique ever since Ogawa et al., (1990) discovered the level of
blood oxygen was visible on certain types of MRI scans referred to as “blood oxygen-level
dependent” (BOLD) contrast. The BOLD technique depends on the fact that deoxygenated
haemoglobin is paramagnetic and becomes magnetic when placed in a magnetic field, while
Neuroeconomics of Individual Decision Making at Multiple Levels 163
oxygenated haemoglobin does not exhibit such properties. Ogawa et al., (1992) applied the
BOLD technique to image activity in visual cortex to alternating on/off visual stimulus. The
fMRI technique has relatively high spatial resolution (~1mm) and low temporal resolution
(1sec) (Logothetis et al., 2001; Heeger and Ress, 2002; Ugurbil, Toth and Kim, 2003;
Logothetis, 2008). The disadvantage with fMRI technique is the artefacts in some areas of the
brain near air-filled cavities such as medial orbitofrontal, inferior temporal cortex.
Electrical and haemodynamic techniques have different resolutions of imaging. While
fMRI has high spatial resolution (~ 1 mm) and low temporal resolution (> 1 second),
EEG/MEG has low spatial resolution (~ 1 cm) and high temporal resolution (~ 1 millisecond).
Multimodal imaging allows integration of fMRI and MEG/EEG results to combine the
advantages of these techniques (George et al., 1995; Dale, and Halgren, 2001; Shibasaki,
2008; Bandettini, 2009).
While brain imaging techniques demonstrate the involvement of a brain area during
performance of a behavioural task these techniques cannot tell us whether a particular brain
area is essential for the concerned behaviour (Chandrasekhar, 2005; Miyapuram, 2008).
Lesions to particular brain areas allow determination of whether a particular brain region is
necessary for a particular cognitive function. Lesion studies can be performed in animals to
produce very anatomically precise lesions. In humans, patients with well defined lesions (for
example, due to injury) can be studied, but is less precise due to the varied nature of lesions.
With new technique of Transcranial Magnetic Stimulation (TMS), reversible lesions can be
momentarily induced by passing electrical current in the cortex, thus allowing study function
of human brain at systems level. Similarly, TMS can be used to stimulate specific region of
the brain to elicit activation
Many a time the outcomes are probabilistic and have some degree of uncertainty
associated with them. Pascal, way back in 1670 (Pascal, 1670 /1958), conjectured that human
choice behaviour could be understood by the expected value (product of probability and
objective value i.e. magnitude of the outcome).
For example, consider a situation in which an individual is faced with a choice between
two lotteries A and B. Lottery A has an equal chance (50% chance or probability) of winning
$10 and $90 and lottery B has an equal chance of winning $40 and $60. The expected values
of the two lotteries are
The expected values of both the lotteries A and B are the same. Hence, the individual
must be indifferent to either of these two lotteries. Consider another situation in which the
two lotteries are A: 25% probability of winning $10 and 75% probability of winning $90 and
B: 25% probability of winning $40 and 75% probability of winning $60.
The expected values of the two lotteries are
Hence lottery A is expected to be chosen because it has higher expected value than B.
Consider a third situation in which the individual is again faced with two lotteries A: 50%
probability of winning $10 or $90 and B: getting $50 for sure (100% probability). In this
situation, most people would choose lottery B, even though the two lotteries have the same
expected value. This is referred to as risk aversion, i.e. people always tend to choose certain
outcomes and are aversive to choose the risky or uncertain (probability < 100%) outcome.
Figure 1. A hypothetical utility function depicting the relationship between wealth and utility (modified
from Doya, 2008). The utility function f(r; c) = r/(r+c) where c=1 and r is the reward amount (wealth).
Neuroeconomics of Individual Decision Making at Multiple Levels 165
The certainty equivalent is the guaranteed payoff at which a person is indifferent between
accepting the guaranteed payoff and a higher but uncertain payoff. Additionally, the
difference between the certainty equivalent and the expected value of the lottery is known as
the risk premium, which is the amount that the individual actually ‘pays’ in order to avoid
risk. Choices made by an individual can be classified into risk averse, risk seeking or risk
neutral if the certainty equivalent is less than, greater than or equal to the expected value of
the lottery, respectively.
Daniel Bernoulli (1763/1958) proposed that a mathematical function should be used to
correct the expected value to account for risk aversion. Bernoulli suggested that the subjective
value or the utility that people assign to an outcome depends on the wealth of the assigning
person and grows more slowly than its objective value (magnitude). Intuitively this means
that an offer of $10 has more value (=utility) to somebody whose total wealth is $100 than to
somebody richer, whose total wealth is $100,000. More specifically the principle of
diminishing marginal rate of utility (Bernoulli, 1763/1958) states that the utility of each
additional dollar decreases with increasing wealth. Bernoulli proposed that increase in
magnitude is always accompanied by an increase in the utility, which follows a concave
(more specifically, a logarithmic) function of magnitude (Figure 1). Hence, individuals
behave as to maximise the expected utility, instead of the expected value. Extending the
mathematical notion of expected value (Pascal, 1670 /1958), the expected utility of each
alternative is calculated as the sum of utilities weighted (multiplied) by the associated
probability. The alternative with highest expected utility is chosen.
The Expected value and Expected utility are defined as follows:
Where utility is the subjective value that follows a concave function of magnitude for risk
averse people and a convex function for risk seeking people.
Von Neumann and Morgenstern (1944) formalized the Expected Utility theorem
describing under what conditions (or axioms) preferences can be (numerically) represented
using a mathematical function. This should allow for a cardinal representation of preferences
i.e. allowing quantification of how much an option is preferred over another one. The axioms
viz., completeness, transitivity and continuity establish that an individual behaves rationally.
An implicit assumption in expected utility theorem was independence of lotteries. The
independence axiom was challenged by well-known paradoxes (Allais, 1953; Ellsberg, 1961).
The following is an example of Allais Paradox. Given a choice between two lottery A :
10% chance of winning $5 million and lottery B : an 11% chance of winning $1 million,
people would tend to choose the lottery A which has higher expected value, but their
preferences would reverse when an option of winning $1 million for sure is added to both the
gambles. It looks like lotteries with guaranteed outcomes (high probability) are weighted
differently by people. The Ellsberg paradox suggests that people prefer situations with known
probabilities compared to unknown probabilities. This is known as ambiguity aversion.
Depending on whether the probabilities are known or unknown, the uncertain situation can be
classified as risky or ambiguous, respectively. Consider the example of an urn having a total
of 90 balls of which it is known that 30 balls are red and the remaining 60 are either black or
yellow in an unknown proportion. Consider a choice between two options A: You receive
166 V. S. Chandrasekhar Pammi and Krishna P. Miyapuram
$100 if you draw a red ball and B: You receive $100 if you draw a black ball; most people
prefer to choose A. When the two options are modified such that A : You receive $100 if you
draw a red ball or an yellow ball and B : You receive $100 if you draw a black ball or a
yellow ball, most people reverse their preferences and choose option B. The Allais and
Ellsberg paradoxes challenge the Expected utility theory, which is a normative theory.
The expected utility theory is based on two economic parameters, namely utility
(subjective value) and probability. The third parameter that underlies many decision making
situations is the delay, which can be defined as the waiting time until the outcome is realised.
Many decisions in everyday life result in delayed outcomes. Decisions between outcomes that
occur at different points in time are referred to as intertemporal decisions, and the process of
systematically devaluating outcomes over time is called temporal discounting. Discounted
utility model can be considered equivalent to the expected utility theory in time domain. The
expected utility theory suggested that decision makers choose between options based on
probability weighted sum of utilities. The discounted utility model posits that decisions are
made as a weighted sum of utilities with temporal discount factors as weights. In addition to
maximisation of utility rate, the discounted utility model assumes a constant discount rate
(Samuelson, 1937). Constant discounting ensures that a preference between two delayed
options is only dependent on the time delay between them. The order of preferences should be
preserved at all possible time points (time consistency or stationarity). The exponential
discount function has a constant discount rate and is adopted by discounted utility theory.
The expected utility and discounted utility models provide normative descriptions of
human choice behaviour. Empirical research has identified many systematic deviations from
the ideal economic situations. While the normative models tell us what the optimal decisions
should be, several descriptive models have been proposed to adequately describe human
choice behaviour. Below, we describe two such models – the prospect theory and the regret
theory.
Two key assumptions of expected utility theory are: 1) preferences should not depend on
procedure (the way in which they are elicited), description, and context i.e. the other available
options, and 2) preferences should depend on impact of consequences on final states of
wealth. Empirical research has identified a number of anomalies in choice behaviour. For
example, most people prefer a gamble of winning $3000 for sure over a risky gamble yielding
$4000 with 80% probability. However when the gamble is presented for losses, people have
their preferences reversed and choose the risky option when gambling to avoid losses
(Kahneman and Tversky, 1979). This is known as reflection effect i.e. the risk attitude
reverses because of the reflection of prospects (options) around 0 (changing from gains to
losses). Furthermore, Kahneman and Tversky (1979) demonstrated that the gains must be
much larger than the losses in order that subjects would choose to play a mixed gamble
involving both gains and losses. This is referred to as loss aversion suggesting that individuals
are more sensitive to losses (steeper utility function) than to gains. Loss aversion is closely
linked to the endowment effect, which refers to the tendency of people to value an object they
possess more highly than they would if they did not possess it (Tversky and Kahneman,
1991). Lastly, the way the problem is defined determines the choice of the individual
(Kahneman and Tversky, 1984; Kahneman and Tversky, 2000). For example, assume that a
ship sailing with 600 people is going to be drowned, two options are available A: 200 people
will be saved and B: there is a 1/3 probability that 600 people will be saved and a 2/3
probability that no one will be saved. Most people prefer option A. When the problem was
Neuroeconomics of Individual Decision Making at Multiple Levels 167
framed negatively and people were asked to choose between two options C: 400 people will
die and D: there is 1/3 probability that no one will die and 2/3 probability that 600 people will
die, most people preferred option D. In all the four options, the expected value was identical
(200 lives saved), but the way the problem was framed changed the preferences of people
from risk averse (option A) to risk seeking (option D). This is the well-known framing effect.
The Prospect theory of Kahneman and Tversky (1979) suggests the value function to be
defined in relation to a reference point (usually the current wealth) and follows a concave
function for gains (values above the reference point) and a convex function for values lower
than the reference point (losses). In addition, the function is steeper for losses than for gains.
While the expected utility theory assumes that people weigh outcomes according to the actual
probability of occurring, prospect theory suggests non-linear probability weighting.
Individuals calculate the weighted sum of values of outcomes as a non-linear function of
probability over-weighing small probabilities and under-weighing middle and large
probabilities. According to this theory, the value of a prospect is defined as the product of
value function and the probability weighting function (Figure 2). It is to be noted that the
utility function is substituted with a subjective value function defined over gains and losses
relative to a status quo (reference point) and is weighted by a probability weighting function.
Figure 2. Value and probability weighting functions from prospect theory (adapted from Trepel, Fox
and Poldrack, 2005). (A) Value Function V(.) is plotted as a function of gains and losses with respect to
a reference point. V(q) = qα where q≥0, and V(q) = -λ (-q)β where q<0 (here, α=088, β=0.88, λ=2.25).
The α, β measures the curvature of the value function and λ is the coefficient of loss aversion. (B)
Probability weighting function W(.) for gains plotted as a function of the probability p of occurrence of
an event. W(p)= ζ pγ / (ζ pγ + (1-p) γ) where ζ represent the elevation of weighting function and γ
measures its curvature (here, ζ=0.69, γ=0.69).
168 V. S. Chandrasekhar Pammi and Krishna P. Miyapuram
Figure 3. The exponential, hyperbolic and quasi-hyperbolic temporal discount functions (modified from
Berns, 2007; Engelmann and Brooks, 2009) for 50 time points. A) The exponential discount function
(γt) where γ (here we took, 0.95) is the discount rate and t is the time factor. B) The hyperbolic discount
function (1/(α*t+1)) where α=0.1. C) The Quasi-hyperbolic function is a piece-wise function similar to
exponential discount function except for the first time unit (1, β*δ, β*δ2,…, β*δt) where β and δ are
two-factors modelling the constant discount rate and discount rate in the standard exponential formula
(here, β=0.792, δ= 0.96).
The regret theory is another popular descriptive model of human choice behaviour. An
individual feels regret when he makes a choice that results in an outcome that is worse than
what would have occurred had he made an alternative choice. A feeling of rejoice occurs
when the option chosen yields a more favourable outcome than an alternative decision
(Chandrasekhar et al., 2008). Interest in experimental economics had started with a theory
developed by (Bell, 1982, 1983); Loomes and Sugden, (1982, 1987) which was used to
explain the violations of the classical expected utility theory. The regret theory explained the
Allais paradox (Allais, 1953) with an anticipatory regret signal that could avoid decisions that
would lead to worse outcomes. The regret theory incorporated an anticipatory component
which is emotional in nature (regret-rejoice function) into the expected utility framework. The
experience of regret or rejoice at the outcome measures the amount of utility or disutility that
the subjects experienced while comparing the obtained outcome with the alternative outcomes
that could have been chosen (called as counter factual process). But importantly, the back
propagated (in time) regret signal would be useful for the decision making process. Interest in
behavioural economics started with this anticipatory regret signal which supported irrational
behaviour while addressing the interaction of emotions with decision making under
uncertainty.
Decisions are often encountered with temporal factors. The intertemporal choices are the
decisions about the consequences that play out in time (Engelmann and Brooks, 2009) and are
seen in most of the decisions in our daily lives such as decisions about spending money,
purchase of health insurance, money investment and food intake (Berns, Laibson,
Loewenstein, 2007). This temporal aspect in decision making process involves tradeoffs
Neuroeconomics of Individual Decision Making at Multiple Levels 169
between immediate consumption and considerations about better future payoffs. This
phenomenon was observed as either discounting of positive or negative payoffs or the
preferences of payoffs at different points in time (Frederick, Loewenstein, and O'Donoghue,
2004). Earlier in 1937, Paul Samuelson proposed a model of standard discounted utility
theory explaining how individuals weigh future utilities with the help of exponential
discounting factor (Samuelson, 1937). This model of discounting was popular among
economists investigating choice behaviour (Green and Myerson, 1996) where the present
value is equal to the discounted value of a reward amount (the actual value) available after a
delay (in units of time). Another competing model backed by psychologists was the
hyperbolic discounting model. This model explained people sensitiveness to time delays
associated with prospects and the subjective value of a prospect (in this model) depended on
the ratio of amount to time. Another model consisting of quasi-hyperbolic discount function
incorporating additional features in discounting function (Phelps and Pollack, 1968; Laibson,
1997) was proposed while explaining the choices among smaller, sooner payoffs and larger,
later payoffs. The discount function is a two-parameter function (beta-delta preference) with
beta (value ranging from 0 to 1 and is constant discount for all future time periods) and delta
(value less than or equal to 1 and is discount rate in the standard exponential formula) factors.
The equations along with the forms of discounting functions are shown in Figure 3.
Most of our decisions under uncertainty are either risky or ambiguous. Risky decisions
are taken when the probabilities of uncertain outcomes are known whereas ambiguous
decisions are taken when the probabilities of uncertain outcomes are unknown (O’Neill and
Kobayashi, 2009). Behavioural evidence illustrating the distinction between risk and
ambiguity in decision making was first depicted by Ellsberg (1961). He observed that the
subjects responded differently to risky and ambiguous choices, and they mostly selected
choices with risky outcomes. This demonstrates the existence of ambiguous averse people
who would like to choose uncertain outcomes with only known information.
When people were given a choice between lotteries with sure smaller valued outcome
versus an uncertain outcome with a larger value, some people chose the sure outcome
whereas some selected the uncertainty outcome while risking themselves. The people who
took sure certain outcomes were called risk averse while the people who preferred risky
outcome (preferring a lottery) were called risk seeking. If individuals were indifferent
between a lottery and its expected value, then they are called as risk neutral (Friedman and
Savage, 1948; Kalenscher, 2009). Later, Tversky and Kahneman (1992) elegantly
demonstrated that in low probability gain conditions, subjects were risk seeking, and in low
probability loss conditions they were observed to be risk averse. In addition, under high
probability gain conditions, subjects were risk averse and in high probability loss conditions,
the subjects were observed to be risk seeking. This result explained real life behaviours such
as investing, purchase of insurances, etc.
Every choice under uncertainty could be associated with a risk factor and that would
reflect subjects’ preference over the available options (uncertain in nature). The behaviour of
subjects whether they are risk seeking or risk averse can be measured with the help of risk
factor. The risk factor for a gamble is computed as the variance of the outcome of the selected
options and the difference in risk computed among options could be used in the decision
making (Engelmann and Tamir, 2009).
Risk and ambiguity are important components of decision making. Loewenstein and
colleagues (2001) with risk-as-feelings hypothesis argued that feelings (which was dealt as
170 V. S. Chandrasekhar Pammi and Krishna P. Miyapuram
separate component in the earlier decision making models) influence cognitive evaluations
and also effect behaviour (which occurs after the cognitive evaluation). Their argument
evolved into a framework consisting of bidirectional link between feelings and cognitive
evaluation modules, and further influences behaviour while leading to outcomes.
freely choose between the two targets. It was found that the probability of a response and
activity in the LIP matches the probability that it will be rewarded although this is less
optimal than making the most rewarded response every time. More precisely, the ratio of
choices between two alternatives will be the same as the ratio of probabilities of these
alternatives being rewarded. This is in accordance with the well known matching law
(Herrnstein, 1961). Sugrue et al. (2004) demonstrated that in a dynamic foraging
environment, LIP neurons tracked the changing values of alternative choices through time.
They concluded that matching provides a behavioural method of observing the internal
representation of value. Recently Kiani and Shadlen (2009) demonstrated that the neurons in
the parietal cortex represented the formation of decision direction and also the degree of
certainty underlying the decision that was opted out.
Primate decision making is also reported to involve other areas of the brain such as the
anterior cingulate cortex (Kennerley et al., 2006; Rushworth and Behrens, 2008), posterior
cingulate cortex (McCoy and Platt, 2005), cingulate motor area (Shima and Tanji, 1998),
dorsolateral prefrontal cortex (Lee and Seo, 2007), orbitofrontal cortex (Padoa-Schioppa and
Assad, 2006; Tsujimoto, Genovesio, and Wise, 2009) and basal ganglia (Samejima et al.,
2005; Samejima and Doya, 2007; Lau and Glimcher, 2008).
The basal ganglia, particularly the ventral striatum (and nucleus accumbens) have a
critical role in reward, decision making, and motivated behaviour. Dopamine, a modulatory
neurotransmitter, is linked to reward prediction (Doya, 2008). The dopaminergic system
projects heavily to the striatum and prefrontal cortex. Dopamine neurons fire in response to
unpredicted rewards and stimuli predicting rewards. The observation that dopamine neurons
show suppressed activation from their baseline firing when a predicted reward does not occur
lead to the hypothesis that dopamine neurons represent reward prediction error (Schultz et
al., 1997; Daw and Doya, 2006). Dopamine appears to mediate the motivational aspects of a
stimulus as blocking dopamine impairs wanting but not liking of reward (Berridge and
Robinson, 1998). Schultz and colleagues have convincingly demonstrated that dopamine
neurons represent economic parameters of decision such as magnitude, probability, expected
value and uncertainty (Fiorillo, Tobler, and Schultz, 2003; Fiorillo, 2008; Schultz et al., 2008;
Fiorillo, Newsome, and Schultz, 2008). When monkeys learned stimuli with varying levels of
reward probability, an increase in sustained firing for stimuli with greater risk (closer to 0.5
reward likelihood) was observed during the delay period between stimulus and reward
(Fiorillo et al., 2003).
Using single-cell recordings on pigeons Kalenscher et al. (2005) presented evidence that
the brain region, which is functionally analogous to prefrontal cortex varied inversely with the
length of delay in time during anticipation for reward and co-varied with the expected value
of the outcome. Further this study revealed the reward preference shift from large to smaller
rewards and validated the hyperbolical discount function incorporating amount of reward and
time delay based on neural data obtained. In a recent study on three rhesus monkeys Kim et
al. (2008) found that the prefrontal cortex (especially the dorsolateral prefrontal neurons)
encoded the temporally discounted value of reward expected from a particular choice.
Neurophysiological studies on primates and pigeons have provided ample evidence that
single neurons (and groups of them) can compute decision variables in the brain.
172 V. S. Chandrasekhar Pammi and Krishna P. Miyapuram
According to decision affect theory (Mellers et al., 1997), responses to a given outcome
depend on counterfactual comparisons. Breiter et al. (2001) presented subjects with three
outcomes in which subjects could win or loose money. Three kinds of prospects (good: $10,
$2.50, $0, intermediate: $2.50, $0, -$1.50 and bad: $0, -$1.50, -$6) were used. Thus $0 on a
good prospect will be experienced as a loss and the same outcome in a bad prospect would be
experienced as a win. Partial evidence for this was observed clearly in time courses of nucleus
accumbens and amygdala for the good and bad prospects, but not so for intermediate
prospect. Haemodynamic responses in the amygdala and orbital gyrus tracked the expected
values of the prospects.
DeMartino et al. (2006) found amygdala activity that correlated with the behavioural
prediction of the framing effect (Kahneman and Tversky, 1979); its activity was enhanced
when the choice was safe (as opposed to risky) in the gains domain, whereas the inverse
happened in the losses domain. On the contrary, dorsal anterior cingulate cortex (ACC)
activity followed the opposite pattern (higher activity for the risky option in the gains domain;
higher activity for the safe option in the losses domain). In addition, medial orbitofrontal
cortex (OFC) activity correlated with the susceptibility to framing effect. Poldrack and
colleagues (Tom et al., 2007) tried to measure BOLD responses correlating with loss
aversion. They offered binary (p=0.5) mixed gambles resulting in either a positive or a
negative outcome. Typical areas of the reward system, and especially ventral striatum and
ventromedial prefrontal cortex, were activated in correlation with both increasing gains and
decreasing losses.
The neuroeconomic studies on regret and rejoice using functional MRI consistently
reported medial orbitofrontal cortex (OFC) involvement with the experience of regret
(Camille et al., 2004; Coricelli et al., 2005; Chandrasekhar et al., 2008). The experience of
rejoice or relief signals were found in the cortical and subcortical brain regions such as
ventral striatum and mid-brain (Chandrasekhar et al., 2008), and anterior ventrolateral
prefrontal cortex (Fujiwara et al. 2009).
Decisions could also incorporate disappointment or elation signals. Disappointment is a
reduction in utility from a bad outcome purely due to an unfavorable realization of a random
variable. When the outcome obtained is lower than expected, subjects might feel
disappointment (Loomes and Sugden, 1986; Roese, 1997). Elation is opposite to that of
disappointment and analogous to rejoicing. The experimental conditions of these
neuroeconomic studies (Camille et al., 2004; Coricelli et al., 2005) were designed based on
partial feedback to investigate neural correlates of disappointment. In partial feedback
conditions only the opted gamble output was shown but in their full feedback conditions, the
outcome of decision along with the outcomes of alternatives were shown. Coricelli et al.,
(2005) showed the involvement of midbrain (periaqueductal gray matter), precentral gyrus
(S2), subcallosal gyrus, middle temporal gyrus correlated with the magnitude of
disappointment.
Dread is an important temporal component in the process of decision making. In a
decision making study with the incentives being milder electrical shocks, Gregory Berns and
colleagues at Emory University investigated neurobiological substrates for dread (Berns et al.,
2006) and observed modulation in subjective experience of dread through BOLD responses
measured in the pain network. This study illustrated the importance of functional MRI
methodology to distinguish subjective experience (disutility) of dread while explaining the
temporal aspect in decision making under uncertainty. This temporal aspect called as
174 V. S. Chandrasekhar Pammi and Krishna P. Miyapuram
intertemporal choice often effects decisions (Berns, Laibson, Lowenstein, 2007). Recently,
Carter and Krug (2009) pointed to the role of amygdala and anterior cingulate cortex in the
anxiety related processes (such as dread) and suggested further work in establishing the
circumstances in which the functional activity in anterior cingulate cortex (ACC) for
diagnosis and treatment of dread related disorders.
Camerer and colleagues (Hsu et al., 2005) are one of the first people to show the neural
correlates of risk and ambiguity associated with choices under uncertainty based on
neuroimaging data from normal population and behavioural data from frontal lobe diseased
patients .They found the dorsal striatum was more sensitive to risk than to ambiguity and the
level of ambiguity was positively correlated with the activations in amygdala and lateral
orbitofrontal cortex. Interestingly, they found that the ambiguity aversion correlated with the
activations in the orbitofrontal cortex and behavioural results from patients with orbitofrontal
lesions were risk- and ambiguity- neutral.
In another recent study, Huettel et al. (2006) used a modified paradigm with certain and
uncertain choices to dissociate risk (uncertainty of known probabilities) and ambiguity
(uncertainty of unknown probabilities) related brain regions. They demonstrated the
involvement of lateral prefrontal cortex with ambiguity and the posterior parietal cortex with
risk associated with the decision making process with monetary gambles. In summary, these
two studies (Hsu et al., 2005; Huettel et al., 2006) demonstrated the involvement of frontal
cortex and amygdala with ambiguity, and the posterior partietal cortex and straitum with risk
associated with the decision making process. These studies categorically defined these two
processes (ambiguity and risk) either by fully presenting or completely avoiding one of the
two (O’Neill and Kobayashi, 2009).
In a recent study incorporating partial conditions of ambiguity using aversive outcomes
(electrical shocks), Bach et al. (2009) varied ambiguity with the help of three cues; no
ambiguity (risky cues), intermediate level of ambiguity and full ambiguity (that they called as
ignorance cue). Their results show the involvement of posterior inferior frontal gyrus and
posterior parietal cortex with the intermediate ambiguous cue compared with risky and
ignorance cues. However, they could not observe any activation for risky or ignorance cues
compared with the intermediate ambiguity cues. As an indication, the cognitive processes
such as ambiguity or risk or ignorance is prevalent in any decision making process under
uncertainty and thus should be given a thought while designing any study.
Engelmann and Tamir (2009) used choice between uncertain lotteries to investigate the
neural correlates of subjective valuations involving risk factor. Their results show increased
activations in anterior and posterior cingulate cortex, superior frontal gyrus, caudate nucleus,
and substantia nigra as a function of risk levels. Their further analysis revealed significant
correlations between risk-seeking attitudes and neural activity in superior and inferior frontal
gyri, medial and lateral orbitofrontal cortex, and parahippocampal gyrus. The risk-aversive
attitudes were observed to be correlated with the activations in caudate.
A recent study by Shultz and colleagues distinguished reward value coding from risk
attitude related brain activations (Tobler et al., 2007). In this study they presented stimuli of
varying expected values by manipulating magnitude and probability of rewards. Their results
suggested the role of striatum in combining the magnitude and probability of rewards to
produce expected value signal. Further they show the risk aversion choices were correlated
with responses from the lateral orbitofrontal cortex and the risk seeking choices were sub-
served by medial sites of the orbitofrontal cortex (OFC).
Neuroeconomics of Individual Decision Making at Multiple Levels 175
period. A frontal network co-varied with the reward prediction error signal both at the time of
the cue and at the time of the outcome. The ventral striatum showed sustained activation that
co-varied with maximum reward uncertainty during reward anticipation. Their results suggest
distinct functional networks encoding economic decision parameters.
Berns et al. (2001) delivered subjects with fruit juice and water in a temporally
predictable or unpredictable manner. Unpredictability of rewards resulted in significant
activity in nucleus accumbens and medial orbitofrontal cortex, while predictability resulted in
activation predominantly in the superior temporal gyrus. Unlike reward learning, the source
of prediction in Berns et al (2001) was based on the sequence of stimuli. Reward learning
literature has focused on the computational models such as the temporal difference (TD)
model incorporating temporal prediction error. Using appetitive conditioning paradigm
O’Doherty et al., (2003b) demonstrated activity in the ventral striatum and OFC with the error
signal when taste reward was omitted or unexpectedly delivered in some of the trials. Signals
predicted by the temporal difference models were found to correlate with activity in the
ventral striatum and the anterior insula in a second-order pain learning task (Seymour et al.,
2004). Seymour et al. (2007b) found that striatal activation reflected positively signed
prediction error in anterior region for rewards. However, the activation was found in posterior
regions for losses in a probabilistic Pavlovian task to compare winning / losing money in two
conditions when the alternative was respectively winning / losing nothing, or losing / winning
money. In the next section, we provide an overview of these computational models and
suggest how these can be extended taking economic decision making into consideration.
Loss aversion is an important concept that originated from the prospect theory developed
by Kahneman and Tversky (1979) using hypothetical risky gambles and was explained with
the help of value function varying between losses and gains. They proposed that the value
function was generally concave for gains and convex for losses. They also suggested that the
reference point was important and the deviations (gains or losses) were dependant on this
point. The important observation from their experimental results was that the value function
was steeper for losses compared to gains. It lead to an important concept called loss aversion,
which states that people exhibit greater sensitivity to losses than gains (Trepel, Fox and
Poldrack, 2005) while making decisions under risk (or uncertainty). The loss aversion was
studied experimentally using buying and selling of goods or losses and gains of monetary
gambles. The important point to be noted while considering individual variability in decision
making process was that some people are not loss averse (Abdellaoui, Bleichrodt and
Paraschiv, 2007).
Two studies investigated the neural basis for loss aversion using functional MRI on
normal humans. With the help of buying and selling of goods (MP3 songs), Weber et al.
(2007) demonstrated loss aversion for goods and money. They suggested that activations in
left amygdala and left caudate nucleus are associated with loss aversion for goods (while
selling songs), while activations in right parahippocampal gyrus are associated with loss
aversion for money (while buying songs). Tom et al (2007) recently investigated loss aversion
for monetary gambles. The subjects in their study either accepted or rejected a gamble
offering 50/50 chance of winning/losing monetary value. Their results showed that the
activity in a unique set of brain regions such as ventral striatum and ventromedial prefrontal
cortex increased with gains and decreased with losses. Based on the slope of value function
for losses, they further demonstrated that these two regions exhibit neural correlates of
behavioural loss aversion (Dreher, 2007). These studies leave us with an open question as to
Neuroeconomics of Individual Decision Making at Multiple Levels 177
whether loss aversion is due to framing effect of the experiment or is medium (goods or
money) dependant.
The following recursive relationship allows estimation of the current prediction and
avoids the necessity to wait until all future rewards are received in that trial.
We can now define the temporal difference error that must approach zero with learning as
The temporal difference model forms the class of computational models referred to as
reinforcement learning. Implementations of reinforcement learning models such as the actor-
critic architecture provide an account of choice behaviour. The actor/critic model has two
components – Actor and Critic. The actor implements a policy for which actions should be
chosen depending upon their predicted outcome. The critic provides an error signal
comparing the predicted outcome to the actual or obtained outcome. An individual learns to
achieve a goal (maximize reward) by navigating through the space of states (making
decisions - actor) using the reinforcement signal (updating the value function - critic). In the
temporal difference (TD) model, the TD error guides the updating of value function V(St)
when transitioning from state St to state St+1. Q-learning and its variants have offered
estimation of value functions over state-action pairs, so that in a given state s, the organism
chooses the action a that maximizes the value Q(s,a). The updating of value function Q is
done similar to the TD model (Watkins and Dayan, 1992).
Research in computational models incorporating the regret signal is becoming important
(Cohen, 2008; Marchiori and Warglien, 2008). Most of the current models use prediction-
error signal for learning the internal parameters of the system. Another class of error signal
that originated from the regret theory, the fictive-error signal, (Chandrasekhar et al., 2008;
Lohrenz et al., 2007) was argued to be essential in future learning algorithms. This error
signal compares the obtained outcome with the expected values of the alternatives that could
have been chosen. This error is different from the prediction error signal which compares
only the obtained outcome with the expected (or desired) outcome.
Several considerations are needed for applying the predictive learning models mentioned
above to economic decision making. First, the future reward value should be replaced by the
expected utility of the future reward incorporating both the magnitude weighting and
probability weighting. Second, a large body of literature has made a clear proposition that
gains and losses are treated differently by individuals. Hence predictive learning models have
to consider the relative weights of gains versus losses instead of a simple flipping of the sign
to represent losses. Third, the temporal discounting function used conforms to the exponential
discounted utility. The more general hyperbolic (or quasi-hyperbolic) discounting function
needs to be incorporated for weighting the time points.
In the following section, we discuss the generalization of individual decision making to
collective decision making.
Most of the research in decision making process in the social context began few years
ago with experimental paradigms involving various components such as altruism, trust,
empathy, cooperation and competition. One of the widely used experimental paradigm was
ultimatum game. The classic version of an ultimatum game first proposed by Guth et al.
(1982) involved two players in which one was a proposer (say, X) and the other was a
receiver (say, Y). The player X is endowed with an amount with which he/she has to make an
economic exchange with his/her peer. The player X (who is also called proposer) would offer
part of the amount endowed to him/her to the receiver Y. The player Y can accept or reject
the offer. If he/she accepts the proposed division amount, the economic exchange gets
implemented as it is. If the player Y rejects the offer, both players get nothing.
There are also experimental paradigms with which the reciprocal interactions among
individuals during economic exchange can be investigated. One example is a trust game, in
which a player called investor will make an interaction with a partner called trustee. The
investor decides how much to endow to invest with trustee and during the transfer to trustee
the amount gets multiplied (say, thrice) and trustee would return some or all amount. This is
an elegant paradigm to investigate the economic investment and exchange with trusted
partners (Rilling, King-Casas, Sanfey, 2008).
It is interesting to note that some components associated with economic exchange such as
envy, revenge, altruism, trust, empathy, cooperation, and competition between two players
can be studied using the principles of these two paradigms. We believe that the modified two
player game paradigms can be utilized to investigate the collective decision making process.
There are also experimental paradigms involving strategic interactions (Camerer, 2004) and
the equilibrium (Nash, 1950).
The results from important neuroeconomic studies using two-person economic games
along with a possible methodology will be reviewed in this section. Montague, Berns et al.
(2002) proposed a new methodology using simultaneous functional MRI called
hyperscanning for recording neurobiological underpinnings during social interactions.
Though there are many technical and conceptual bottlenecks to be solved, in future, this
methodology might play an important role for investigation of collective decision making or
interpersonal economic exchanges.
In one of the first neuroeconomic investigations using ultimatum game, Sanfey and
colleagues (2003) depicted the neural correlates of emotional and cognitive processes. They
found unfair offers activated anterior insula and dorsolateral prefrontal cortex and they further
observed enhanced activity in anterior insula when the unfair offers were rejected by the
responder. A subsequent study by the same group (Wout et al., 2005), using repetitive
transcranial magnetic stimulation technique, demonstrated the causal role of right dorsolateral
prefrontal cortex (DLPFC) in the strategic decision making process.
Using functional MRI with a modified multi-round trust game involving economic
exchange, King-Casas et al. (2005) demonstrated the involvement of caudate (in dorsal
striatum) with reciprocity involving trust. One of the important findings from this study was
that the caudate registered the social prediction errors and eventually guided the reciprocity.
Using a computer game to investigate the neural correlates of cooperation and competition,
Decety et al. (2004) demonstrated the involvement of medial orbitofrontal cortex in the
process of cooperation and that of inferior parietal and medial prefrontal cortices in the
process of competition. Using an intranasal Oxytocin infusion (a neuropeptide) on human
subjects playing a trust game, Ernst Fehr and colleagues (Kosfeld, 2005) demonstrated that
180 V. S. Chandrasekhar Pammi and Krishna P. Miyapuram
Oxytocin has a role in increasing initial money transfers by the investors and thus point to its
key function in economic interaction with trust. A recent study by Montague and colleagues
(Tomlin et al., 2006) investigated two-person economic exchange and found the role of
cingulate cortex in agent-specific responses. Using emotional faces to demonstrate empathy,
Carr et al. (2003) showed the important role played by the brain region Insula.
Though there are many studies in the literature, we have pointed to some important ones.
The questions that remain open to neuroimaging methods are how the network activity, the
connectivity among brain regions associated with the economic exchange, would predict the
nature of interaction and how this would influence the partners in the economic exchange.
The methodological issues related to hyperscanning should also be addressed.
7. CONCLUSIONS
In this paper, efforts were made to bring together various components of neuroeconomics
of decision making under uncertainty. Methods for investigating brain functions, the various
economic theories of behaviour of human choice, the neuronal mechanisms sub-serving at the
cellular and the system levels, and the computational models relevant to the process of
decision making were reviewed. The future research in collective decision making along with
some open questions related to neuroeconomics were also presented.
ACKNOWLEDGMENTS
We are thankful for inspirational support from our current and past colleagues and
mentors. We thank Drs. Raju Bapi, Kenji Doya and Ahmed. VSC Pammi is thankful to
Narayanan Srinivasan, Ranganatha Sitaram, Debarati Banerjee, Gregory Berns, Giuseppe
Pagnoni, Charles Noussair, Monica Capra, Jan Engelmann and Sara Moore. KPM would like
to thank Yorgos Christopoulos, Philippe Tobler, Shunsuke Kobayashi and Wolfram Schultz.
REFERENCES
Abdellaoui, M., Bleichrodt, H., and Paraschiv, C. (2007). Loss aversion under prospect
theory: a parameter-free measurement. Management Science, 53(10), 1659-1674.
Allais, M. (1953). Le comportement de I’Homme rationnel devant le Risque : Critique des
tostulats et axiomes de I’Ecole Americaine. Econometrica, 21, 503-546.
Bach, D. R., Seymour, B., and Dolan, R. J. (2009). Neural activity associated with the passive
prediction of ambiguity and risk for aversive events. The Journal of Neuroscience, 29,
1648-1656.
Bell, D.E., 1983. Risk Premiums for Decision Regret. Management Science, 29, 1156-1166.
Bell, D.E., 1982. Regret in decision making under uncertainty. Operations Research, 30, 961-
981.
Bandettini, P. A. (2009). What's new in neuroimaging methods? Annals of the New York
Academy of Sciences, 1156, 260-293.
Neuroeconomics of Individual Decision Making at Multiple Levels 181
Berns, G. S., McClure, S. M., Pagnoni, G., and Montague, P. R. (2001). Predictability
modulates human brain response to reward. The Journal of Neuroscience, 21, 2793–
2798.
Berns G. S., Chappelow, J., Cekic, M., Zink, C. F., Pagnoni, G., and Martin-Skurski, M. E.
(2006). Neurobiological substrates of dread. Science, 312(5774), 754-758.
Berns, G. S., Laibson, D., and Loewenstein, G. (2007). Intertemporal choice--toward an
integrative framework. Trends in Cognitive Science, 11(11), 482-488.
Berns, G.S. (2008). Neuropolicy Center Confronts the Biological Basis of Collective
Decision Making. Press report, Emory University, Atlanta, USA.
Berridge, K. C., and Robinson, T. E. (1998). What is the role of dopamine in reward: hedonic
impact, reward learning, or incentive salience? Brain Research Reviews, 28(3), 309-369.
Bernoulli, D. (1763/1958). Exposition of a new theory on the measurement of risk.
Econometrica, 22, 23-36.
Boettiger, C. A., Mitchell, J. M., Tavares, V. C., Robertson, M., Joslyn, G., D'Esposito, M.,
and Fields, H. L. (2007). Immediate reward bias in humans: Fronto-Parietal networks and
a role for the Catechol-O-Methyltransferase 158Val/Val genotype. The Journal of
Neuroscience, 27(52), 14383–14391.
Breiter, H. C., Aharon, I., Kahneman, D., Dale, A., Shizgal, P. (2001). Functional imaging of
neural responses to expectancy and experience of monetary gains and losses. Neuron, 30,
619–639.
Camerer, C.F. (2004). Behavioral game theory: Predicting human behavior in strategic
interactions. In C. F. Camerer, G. Loewenstein and M. Rubin (Eds.), Advances in
Behavioral Economics (pp. 374-292). New York: Princeton University Press.
Camille, N., Coricelli, G., Sallet, J., Pradat-Diehl, P., Duhamel, J. R., and Sirigu, A. (2004).
The involvement of the orbitofrontal cortex in the experience of regret. Science, 304,
1167-1170.
Carr, L., Iacoboni, M., Dubeau, M-C., Mazziotta, J. C., and Lenzi, G. L. (2003). Neural
mechanisms of empathy in humans: A relay from neural systems for imitation to limbic
areas. Proceedings of National Academy of Sciences, USA, 100, 5497-5502.
Carter, C. S., and Krug, M. K. (2009). The functional neuroanatomy of dread: Functional
magnetic resonance imaging insights into generalized anxiety disorder and its treatment.
American Journal of Psychiatry, 166(3), 263-265.
Chandrasekhar, PVS. (2005). Functional MRI investigation of Complex Sequence Learning.
PhD Thesis, Department of Computer and Information Sciences, University of
Hyderabad, India.
Chandrasekhar, P. V. S., Capra, C. M., Moore, S., Noussair, C., and Berns, G. S. (2008).
Neurobiological regret and rejoice functions for aversive outcomes. Neuroimage, 39(3),
1472-1484.
Churchland, P.S., and Sejnowski, T. J. (1992). The Computational Brain. Bradford Book,
MIT Press, Cambridge MA.
Cohen, M. D. (2008). Learning with Regret. Science, 319, 1052-1053.
Coricelli, G., Critchley, H. D., Joffily, M., O'Doherty, J. P., Sirigu, A., and Dolan, R. J.
(2005). Regret and its avoidance: A neuroimaging study of choice behavior. Nature
Neuroscience, 8, 1255-1262.
Dale, A. M., and Halgren, E. (2001). Spatiotemporal mapping of brain activity by integration
of multiple imaging modalities. Current Opinion in Neurobiology, 11(2), 202-208.
182 V. S. Chandrasekhar Pammi and Krishna P. Miyapuram
Krain, A.L., Wilson, A. M., Arbuckle, R., Castellanos, F. X., and Milham, M. P. (2006).
Distinct neural mechanisms of risk and ambiguity: a meta-analysis of decision-making.
NeuroImage, 32(1), 477-484.
Laibson, D., (1997). Golden eggs and hyperbolic discounting. Quarterly Journal of
Economics, 112, 443–477.
Lau, B., and Glimcher, P. W. (2008). Value Representations in the Primate Striatum during
Matching Behavior. Neuron, 58(3), 451-463.
Lee, D., Seo, H. (2007). Mechanisms of reinforcement learning and decision making in the
primate dorsolateral prefrontal cortex. Annals of the New York Academy of Sciences,
1104, 108-122.
Loewenstein, G. F., Weber, E. U., Hsee, C. K., and Welch, N. (2001). Risk as feelings.
Psychological Bulletin, 127(2), 267-286.
Loewenstein, G. (1987). Anticipation and the valuation of delayed consumption. The
Economic Journal, 97, 666-684.
Logothetis, N. K. (2008). What we can do and what we cannot do with fMRI. Nature,
453(7197), 869-878.
Logothetis, N. K., Pauls, J., Augath, M., Trinath, T., and Oeltermann, A. (2001).
Neurophysiological investigation of the basis of the fMRI signal. Nature, 412, 150-157.
Lohrenz, T., McCabe, K., Camerer, C. F., and Montague, P. R. (2007). Neural signature of
fictive learning signals in a sequential investment task. Proceedings of the National
Academy of Science, USA, 104(22), 9493-9498.
Loomes, G., and Sugden, R., 1982. Regret theory: an alternative theory of rational choice
under uncertainty. The Economic Journal, 92, 805-824.
Loomes, G., and Sugden, R., 1986. Disappointment and dynamic consistency in choice under
uncertainty. The Review of Economic Studies, 53, 271–282.
Loomes, G., and Sugden, R., 1987. Some implications of a more general form of regret
theory. Journal of Economic Theory, 41, 270-287.
Luhmann,C. L., Chun, M. M., Yi, D-J., Lee, D., and Wang, X-J. (2008). Neural dissociation
of delay and uncertainty in Intertemporal choice. The Journal of Neuroscience, 28(53),
14459-14466.
Marchiori, D., and Warglien, M. (2008). Predicting human interactive learning by regret-
driven neural networks. Science, 319, 1111-1113.
McClure, S. M., Laibson, D. I., Loewenstein, G., and Cohen, J. D. (2004). Separate neural
systems value immediate and delayed monetary rewards. Science, 306(5695), 503-507.
McClure, S.M., Ericson, K.M., Laibson, D.I., Loewenstein, G., and Cohen, J.D. (2007). Time
discounting for primary rewards. The Journal of Neuroscience, 27, 5796-5804.
McCoy, A. N., Platt, M. L. (2005).Risk-sensitive neurons in macaque posterior cingulate
cortex. Nature Neuroscience, 8(9), 1220-1227.
Mellers, B.A., Schwartz, A., Ho, K., and Ritov, I. (1997). Decision affect theory: emotional
reactions to the outcomes of risky options. Psychological Science, 8, 423–429.
Miyapuram, K.P. (2008) Human neuroimaging of visual presentation and imagination of
reward. Ph.D. Thesis, University of Cambridge, UK.
Montague, P.R., Berns, G.S., Cohen, J.D., McClure, S.M., Pagnoni, G., Dhamala, M., Wiest,
M.C., Karpov, I., King, R.D., Apple, N., and Fisher, R.E. (2002). Hyperscanning:
Simultaneous fMRI during linked social interactions. Neuroimage, 16, 1159-1164.
Neuroeconomics of Individual Decision Making at Multiple Levels 185
Sanfey, A. G., Rilling, J. K., Aronson, J. A., Nystrom, L. E., and Cohen, J. D. (2003). The
neural basis of economic decision-making in the Ultimatum Game. Science, 300(5626),
1755-1758.
Shibasaki, H.(2008). Human brain mapping: hemodynamic response and electrophysiology.
Clinical Neurophysiology, 119(4), 731-743.
Shima, K., and Tanji, J. (1998). Role for Cingulate Motor Area Cells in Voluntary Movement
Selection Based on Reward. Science, 282(5392), 1335 – 1338.
Schultz, W., Preuschoff, K., Camerer, C., Hsu, M., Fiorillo, C. D., Tobler, P. N., and
Bossaerts, P. (2008). Explicit neural signals reflecting reward uncertainty. Philosophical
Transactions of the Royal Society B: Biological Sciences, 363(1511), 3801-11.
Schultz, W., Dayan, P., and Montague, P. R. (1997). A Neural Substrate of Prediction and
Reward. Science, 275(5306), 1593 – 1599.
Sugrue, L. P., Corrado, G. S., and Newsome, W. T. (2004). Matching behavior and the
representation of value in the parietal cortex. Science, 304(5678), 1782-1787.
Seymour, B., O'Doherty, J.P., Dayan, P., Koltzenburg, M., Jones, A., Dolan, D., Friston, K.
J., and Frackowiak, R. S. J. (2004). Temporal difference models describe higher-order
learning in humans. Nature, 429, 664-667.
Seymour, B., Singer, T., and Dolan, R. (2007a). The neurobiology of punishment. Nature
Reviews: Neuroscience, 8, 300-311.
Seymour, B., Daw, N., Dayan, P., Singer, T., and Dolan, R. (2007b). Differential encoding of
losses and gains in the human striatum. The Journal of Neuroscience, 27(18), 4826-4831.
Sutton, R.S., and Barto, A.G. (1981). Toward a modern theory of adaptive networks:
expectation and prediction. Psychological Review, 88, 135–170.
Sutton, R.S., and Barto, A. G. (1990). Time-derivative models of Pavlovian reinforcement.
In: Learning and computational neuroscience: foundations of adaptive networks. In
Gabriel M, Moore J, (Eds.), pp 497–537. Cambridge, MA: MIT.
Tobler, P. N., O’Doherty, J. P., Dolan, R. J., and Shultz, W. (2007). Reward value coding
distinct from risk-attitude related uncertainty coding in human reward systems. The
Journal of Neurophysiology, 97, 1621-1632.
Tom, S. M., Fox, C. R., Trepel, C., and Poldrack, R. A. (2007). The neural basis of loss
aversion in decision-making under risk. Science, 315, 515-518.
Tomlin, D., Kayali, A. M., King-Casas, B., Anen, C., Camerer, C. F., Quartz, S. R., and
Montague, P. R. (2006). Agent-specific responses in the cingulate cortex during
economic exchanges. Science, 312(5776), 1047-1050.
Trepel, C., Fox, C. R., and Poldrack, R. A. (2005). Prospect theory on the brain? Towards a
cognitive neuroscience of decision under risk. Cognitive Brain Research, 23, 34-50.
Tsujimoto, S., Genovesio, A., and Wise, S. P. (2009). Monkey orbitofrontal cortex encodes
response choices near feedback time. The Journal of Neuroscience, 29(8), 2569-74.
Tversky, A., and Kahneman, D. (1992). Advances in prospect theory: cumulative
representation of uncertainty. Journal of Risk and Uncertainty, 5, 297-323.
Tversky, A. and Kahneman, D. (1991). Loss Aversion in Riskless Choice: A Reference
Dependent Model. Quarterly Journal of Economics, 106, 1039-1061.
Ugurbil, K., Toth, L., and Kim, D. S. (2003). How accurate is magnetic resonance imaging of
brain function? Trends in Neuroscience, 26, 108–114.
Von neumann, J., and Morgenstern, O. (1944). Theory of games and Economic behavior.
Princeton university press. Princeton.
Neuroeconomics of Individual Decision Making at Multiple Levels 187
Volkow, N. D., Rosen, B., and Farde, L. (1997). Imaging the living human brain: Magnetic
resonance imaging and positron emission tomography. Proceedings of National Academy
of Sciences, USA, 94, 2787–2788.
Watkins, C. J. C. H., and Dayan, P. (1992). Q-learning, Machine Learning, 8, 279-292.
Weber, B., Aholt, A., Neuhaus, C., Trautner, P., Elger, C. E., and Teichert, T. (2007). Neural
evidence for reference-dependence in real-marker-transactions. NeuroImage, 35, 441-
447.
Wout, M. V., Kahn, R. S., Sanfey, A. G., and Alerman, A. (2005). Repetitive transcranial
magnetic stimulation over the right dorsolateral prefrontal cortex affects strategic
decision-making. Neuroreport, 16, 1849-1852.
Xue, G., Lu, Z., Levin, I. P., Weller, J. A., Li, X., and Bechara, A. (2009). Functional
dissociations of risk and reward processing in the medial prefrontal cortex. Cerebral
Cortex, 19(5), 1019-27.